The Effects of Rate Regulation on the
Volatility of Auto Insurance Prices
Evidence from Canada
Darrell Leadbetter
Jane Voll
Erica Wieder
________________________
Abstract
Previous studies using U.S. data have found that rate regulation reduces competition,
availability of coverage and increases volatility of insurance premiums. This article
extends the U.S. literature to the Canadian context to examine whether rate regulation
increases premium volatility in the province of Ontario. Based on an empirical
analysis using data covering six provinces over the 18–year period from 1984 to 2001
we find that rate regulation does make insurance premium more volatile for
consumers. This finding is consistent with results from other jurisdictions.
Keywords: price regulation, auto insurance, price volatility
JEL code: L510, G280, L500
Introduction∗
The effect of rate regulation on auto insurance premiums has been the subject of wide
debate and numerous studies. Most empirical research on this question has been
conducted in the United States, taking advantage of the rich heterogeneity of systems of
rate regulation across fifty states. Considering that the Canadian experience with rate
regulation has been somewhat different from that of the United States, the objective of
this study is to extend this literature to Canada and estimate whether rate regulation
increases premium volatility. Intuitively, volatility conjures visions of “choppy” markets
or wide price swings. Throughout this paper we define volatility as a measure of the
degree of price movement in insurance premiums.
American experience with auto-insurance regulation differs from Canada because the
United States McCarran-Ferguson Act of 1945 exempted from federal anti-trust laws any
insurance company that was subject to other state regulations from federal anti-trust laws.
In order to take advantage of this exemption, all states established state regulation of
insurance by 1951. Since 1945, rate regulation in the United States has become a
common feature among many insurance systems in the United States.
Unlike the United States, there has not been the McCarran-Ferguson like catalyst
legislation to encourage government regulation of insurance premiums in Canada. As a
result, active price regulation in Canada is a relatively recent phenomenon and experience
is restricted to Ontario. Prior to 1989, Ontario’s automobile insurance system operated
under a competitive rating model. However, Ontario implemented a strict prior approval
regime of rate regulation following the introduction and passage of Bill 10, the
Automobile Insurance Rates Control Act 1989 and subsequent modifications under Bill
68, the Insurance Law Statute Amendment Act 1990. In 2000, the Financial Services
Commission of Ontario introduced a respond to market (R2M) rate process whereby
filing requirements and approval times are streamlined if rate increases fall below a
threshold determined by the regulator. This change has introduced some limited
flexibility into the rate regulation process.
Among the other regions, Alberta and the four Atlantic Provinces maintain variations on
the file and use system where automobile insurers are required to file rates and, after a
period defined in legislation, acquire “deemed approval” for use. Regulatory authorities
may disapprove a rate filing at any time prior to the “deemed approval” or may extend
the period of evaluation.1 British Columbia currently does not regulate rates for
competitively delivered optional auto coverages. Quebec maintains a use and file system
for private insurers. Overall, automobile insurance remains the only line of insurance
where rates are regulated in Canada.
Rate regulation is a recent phenomenon in competitive markets. The provinces of British
Columbia, Saskatchewan and Manitoba have monopolistic government run insurers for
∗
The authors wish to thank its two reviewers for reviewing the paper and offering suggestions for
improvement. Nevertheless, the authors remain exclusively responsible for all the statements of fact and of
opinion in this paper.
1
Rate filing and underwriting rules were being reviewed in 2003 in Alberta and Atlantic Canada.
1
mandatory basic automobile coverages and these government run insurers have been
subject to state governed price regulation.
The Effects of Rate Regulation
There is a large literature providing statistical evidence of the effects of rate regulation on
average automobile insurance premium levels. A smaller body of research has
investigated the effects of rate regulation on rate volatility and insurance availability.
The customary approach classifies jurisdictions by the form of rate regulation - prior
approval, competitive rating, file and use or some other form – and then compares the
performance between groups controlling for one or more variables.
The current state of research suggests that the effect on rates from prior approval
regulation, relative to claims costs, varies over time (Harrington, 1987; Tennyson, 1997,
& Cummins et al, 2001). The research in the United States has noted that while rate
regulation tends to compress the premiums collected per dollar of loss experience,
insurance premiums in jurisdictions with rate regulation are often higher than in
jurisdictions with less rate regulation (Tennyson, 1997). In the long run rate regulation
has not been found to result in lower prices (Cummins et al, 2001 & Harrington, 2001). It
is interesting to note that seven of the top ten, including all of the top five states with the
highest average premiums over the period 1997 – 2000 actively regulated automobile
insurance rates.
In general, rate regulation has been found to limit competition, reduce availability of
coverage and increase volatility in insurance premiums (Tennyson, 1991 & Harrington,
2002).
Increased volatility in insurance premiums (the focus of this study) could be the result of
delays in the rate approval process under prior approval rate regulation. Regulatory lags
under prior approval rate regulation could produce lower rate increases during periods of
rapid cost growth and larger rate increases or a slower rate of reduction in periods of
stable or declining claims costs (Harrington, 2002). These lags may be the result of the
normal process of regulators working through rate filings or the result of a regulatory
build up. Regulatory build-ups occur where insurers hold off filing smaller, more
frequent, rate increases in favour of larger rate increases that justify the costs of
assembling the detailed actuarial filing requirements.
Empirical analysis of the effects of rate regulation on premium volatility suggests that
active price controls on insurance affect both the amplitude and length of the insurance
underwriting cycle. Analyses of the effects of rate regulation on loss ratios have
provided evidence that active price controls on insurance exacerbates premium volatility
in the United States (Witt and Miller, 1981; Outreville, 1990 and Harrington, 2001). In
addition, there is some cross-country evidence that rate regulation increases the length of
underwriting cycles in the insurance system (Lamm-Tennant and Weiss, 1997, Leng et al
2002). For example, in the United States during the period of 1950 to 1970, automobile
insurance went through three underwriting cycles while all lines except automobile had
only one full cycle. This trend continued during the 1980’s but was moderated by
2
increased competition. The cycle in automobile insurance has been largely statistical,
reflecting regulatory lag in adjusting prices to costs (Cummins et al, 1991).
The purpose of this research is to further investigate the impact of rate regulation on
insurance premium volatility, drawing upon the methodology and experience of the
literature in the United States. To do so, this paper undertakes an econometric analysis of
the effects of prior approval rate regulation of auto insurance in Canada. We then
compare premium volatility in Ontario under its prior approval rate regulation regime
with that of a simulated competitive rating system. Finally, we identify and review two
case studies from the United States that offer a practical and real world test of theoretical
and statistical literature.
The Empirical Model
The analysis presented here draws upon the methodology and approach outlined in
Harrington (2001). In that paper the effects of rate regulation on volatility in the
unexplained growth rate of average premiums is analized. The unexplained growth of
automobile insurance premiums is the growth in auto premiums that is not predicted by
growth in claims costs, accident frequency or other variables expected to contribute to the
cost of insurance. Under this framework:
∆premium = ∆explanatory variables + residual
where the residual represents the unexplained volatility in the system.
Two steps are required. In the first step, we regress average expenditure (total
premiums/insured vehicles) on a set of explanatory variables. In the second step, the
volatility of average expenditure ,estimated from the first step, is then used as a proxy for
the volatility of insurance rates and regressed on a regulation index and a second set of
explanatory variables. To achieve this, we apply panel data regression techniques. Note
the terms, average expenditure and average premium are used interchangeable
throughout.
Model specification:
In the first step, we estimate the following model:
Average expenditurejt = (α + λj) + βjXjt + εjt
where the subscripts j and t refer to province j in year
t, α is the intercept and λj is the unobserved fixed
effect. Xjt is a vector of control variables that could
influence average expenditures apart from any effects
of regulation and βj is the set of coefficients for those
control variables. Since, automobile insurance is a
mechanism for spreading risk, pooling the resources of
many to share the losses of a few, claims costs are
expected to be an important factor in determining
(1)
Table 1: Variable Definitions
Description
Xjt
Average claims
Loss ratio
Herfindahl index
CPI
Average exposure
Total losses/# of claims
Incurred
Losses/premiums
Measure of competition
Consumer price index
Accident frequency
(# of claims / # insured
vehicles insured)
3
insurance expenditures. Therefore, average claims costs are included as a control variable
in order to control for different product/benefit levels among provinces. Similarly, loss
ratios are highly correlated with profitability measures and are included to control for
underwriting capacity and profitability (Harrington, 2002). A measure of competition,
the Herfindahl index, is included to condition the results on the level of competition in
the industry. As the cost of inputs to insurance and claims would be expected to increase
as the general price level rises, the CPI is also included in the set of control variables.
Finally, as motor vehicle collisions are the primary source of claims in automobile
insurance, a control for accident frequency is included.
λj represents the fixed effects to be estimated, and εjt is a statistical disturbance term that
reflects other factors that are not only particular to the individual provinces but also to
time periods. With no further constraints, λj and εjt do not have a unique solution. Thus,
before equation (1) can be estimated, we must place an additional constraint on the
system. In this case the software package we used for the study assumes Σλj =0.
In the second step, using the residuals from equation (1) as a proxy for the volatility of
insurance rates the following model is estimated:
unexplained volatilityjt = (εjt - εjt )2 = (η +Ωj)+ γregulationjt + ФHjt + φjt
(2)
and the assumption of ΣΩj =0 holds. Unexplained volatility is the demeaned residuals of
equation (1), regulation is an index of regulation, Hjt is a vector of control variables, and
φjt is a statistical disturbance term. Our model assumes that sources of unexplained
volatility are the result of structural changes in the environment. Traditionally such
structural changes may include the removal/addition of barriers to competition or a
systemic shock to the system such as the development of a new risk such as terrorism.
We therefore include changes in claims costs and change in the competitive environment
(Herfindahl) as control variables.
In estimating the above equations, we utilized several statistical tests in order to identify
the appropriate estimation technique. Typically, the literature covering panel data
estimation methods makes a distinction between fixed- and random-effects models. Fixed
effects models are usually applied when the λj and Ωj are assumed to be fixed parameters
to be estimated and the remainder disturbances of the models are assumed to be
independent and identically distributed IDD(0,σ2 ). Essentially, using fixed effects is the
same as assuming different intercepts for each province but equal slopes. This model is
appropriately used when we intend to study the behavior of a set of groups and our
inferences are restricted to the behavior of that set only (Baltagi, 2002). Instead, if we
assume that λj and Ωj are random variables (which happens when we consider the
observations to be random draws from a large population), we would use a random
effects model.
The Hausman specification test was performed to determine whether a random
coefficients model would be more appropriate than one with fixed coefficients. Under the
null hypothesis both random effects and fixed effects estimates were consistent and H0:
4
E(ujt/Xjt) =0, when H0 is false, and β̂ RE was inconsistent. The statistic is Chi2=125.07 and
since Prob> Chi2=0 we reject the null hypothesis. This statistical examination suggested
that a fixed effects model was the appropriate specification for our econometric analysis.
In addition, we performed the Breusch-Pagan test for random effects, where the null
hypothesis is that there is no within-unit correlation (and that we should use random
effects). When the null hypothesis of this test is rejected, it indicates that there are
individual specific effects in the model. In this case, the reported statistic is Chi2=49.45,
with Prob> Chi2=0. We therefore reject the null hypothesis as there is evidence that there
are individual specific effects confirming that a fixed effects model is the appropriate
estimation model.
In this study, the results presented are obtained from a fixed effect model, except when
correcting for heteroskedasticity, where we estimate the model with corrected standard
errors using a feasible generalized least squares (GLS) method. The results of the
econometric analysis employing this model were then used to compare premium
volatility in Ontario under the historical prior approval rate regulation with a simulated
file and use scenario.
In addition, several diagnostic tests were performed. We used the Chow test to test the
null hypothesis of whether slopes and intercepts are constant across provinces. The
following equation specifies this test:
F((N-1)K’, N(T-K’)) = (SSRR- SSRUR) / (N-1)K’ = 34.90
SSRUR
(3)
/ N(T-K’)
Based on the results of the Chow test, we rejected the null hypothesis at a 5% level
(F(36,84)=1.59). However, this test treats the residuals as being homoskedastic and
Toyoda (1974) has demonstrated that it is wrong to apply the Chow test in case of
heteroskedastic variances. Therefore, to test for panel level heteroskedasticity we
performed a Likelihood-ratio (LR) test, with the null hypothesis of homoskedasticity. The
statistic is an LR Chi2=69.05, and with a Prob> Chi2=0, we reject the null hypothesis,
indicating the presence of heteroskedasticity. However, this should not present a material
issue in our estimations since they continue being consistent, albeit with efficiency loss.
Nevertheless, we estimated the model using a feasible generalized least squares method
which produces an estimate of the model correcting for heteroskedasticity and compared
the results with the fixed effects model.
Data
The data used in this study were obtained from provincial statistical plans to which
insurance companies submit auto insurance data relating to premiums and losses by type
of coverage.
5
Observations were pooled across six provinces (Ontario, Alberta, Newfoundland, Nova
Scotia, New Brunswick and PEI) over an 18-year period (from 1984 to 2001). The initial
sample contained 80 companies, including high-risk reinsurers or ‘facility’ carriers. The
final sample contained 68 companies excluding 10 facility carriers as well as two
companies for which sufficient data were not available.
Table 2: Type of Rating Regime by Province
To measure rate regulation, we used
provincial
statutory
regulatory
Regime
Jurisdiction
requirements for automobile insurance
No regulation
0 Ontario pre 1989
Advisory/no file
1
rates for insurers, based on data from
Use
and
file
2
Quebec (optional coverages)
provincial regulatory bulletins and
File
and
use
3
Alberta, Atlantic provinces,
industry sources. Two methods were
Quebec (mandatory
used to measure the effects of rate
coverages)
regulation. The first was a dummy
Flex rating
4
variable that took on the value of 1 for
Modified prior approval
5 Ontario 1997 - present
Prior approval
6 Ontario, 1989 – 1997
prior approval regimes and 0 for other
Government determined
7 British Columbia, Manitoba,
forms of rate regulation. The second
rates
Saskatchewan (mandatory
method involved mapping provincial
coverages)
regulatory requirements to the ranking
system widely employed within the literature (Devlin, 2002 & Harrington, 2002). Under
this ranking system, the variable regulation takes on one of eight possible values, ranging
from 0 (no regulation) to 7 (state determined rates).
Once the data was pooled, seven databases were created using aggregated data: one
including all types of coverage, four databases with data by type of coverage (liability,
accident benefits, collision, comprehensive), and two databases containing data for
mandatory (pooled accident benefits and liability) and optional (pooled collision and
comprehensive) coverage.
Premium Volatility and Rate Regulation2
In order to test whether volatility in the unexplained growth rate in the average
automobile insurance premium differs between Ontario’s prior approval rate regime and
the file and use rating systems of five other provinces, we first regressed a vector of
control variables on average premium expenditure as the dependent variable. Various
specifications of the X vector of the model were estimated. Measures of income were
added to the regression to capture any income effects that might lead to higher levels of
insurance expenditure. However they were not significant and did not contribute to the
explanatory power of the model. Measures of household income were therefore excluded
from the final specification of the model. In addition, we added a number of interaction
2
This study estimates the conditional effect of rate regulation on the volatility of the average level of
automobile insurance premium expenditures. It does not consider an interesting and important related
question of whether rate regulation distorts consumer and insurer incentives for loss control and therefore
increases claim costs and average rate levels. However, recent research on the specific effects of regulating
the underwriting (one aspect of rate regulation) of automobile insurance in British Columbia suggests that
such distortionary effects increase accident frequency and claims costs (Kovacs et al, 2004).
6
terms including the interaction between average claims costs and accident frequency and
the interaction between the loss ratio and the Herfindahl. These interaction terms did not
typically contribute much to the overall explanatory power of the model and had no effect
on the overall significance or sign of the regulatory variable in the second state of
estimation. Therefore for parsimony the final specification of the model includes the set
of explanatory variables previously described.
The results from estimating this specification using pooled ordinary least squares (OLS)
and generalized least squares (GLS) methods are reported in Table 3.
Table 3: Coefficients and P-values for Equation (1)
Variable
Average
claims
Loss ratio
Price level
Herfindahl
Accident
frequency
Constant
R2
Within
Between
Overall
P>| t |
0.000
Fixed Effects
Coefficient
Predicted sign
0.109
Positive
P>| t |
0.000
GLS
Coefficient
0.141
Predicted sign
Positive
0.000
0.000
0.668
0.000
-364.166
2.116
-0.013
1605.064
Uncertain
Positive
Negative
Positive
0.000
0.000
0.000
0.000
-553.499
1.845
-0.098
2542.978
Uncertain
Positive
Negative
Positive
0.161
55.633
No
0.370
34.101
No
0.9586
0.9776
0.9182
The direction and significance of the estimated coefficients on the X vector of variables,
with the exception of Herfindahl variable, in the fixed effects model are statistically
significant and of the expected sign. Under GLS estimation the Herfindahl variable is
also significant and of the expected direction. A higher Herfindahl score implies less
competition and theoretically reduced downward pressure on the price of insurance,
therefore we expect a negative sign. Thus the efficiency loss of the fixed effects estimates
as a result of heteroskedasticity falsely reject competition as a factor in the determination
of insurance prices. However, in the heteroskedasticity corrected GLS model,
competition is a statistically significant variable in explaining average premiums and the
negative sign confirms that it exerts a downward pressure on insurance expenditures.
Consistent with intuitive expectations, we find that the average costs of claims and
accident frequency are the primary determinants of insurance expenditures. Together
these two variables account for more than 88% of the explanatory power of the model,
with average claims costs alone accounting for more than three quarters of the model’s
explanatory power. Also, as would be expected, these results suggest that a rising
general price level also contributes to increased insurance premiums.
From Table 3, the sign of the estimated coefficient on loss ratios suggests that average
automobile expenditures are negatively related to this variable. This would appear
counterintuitive as it might be expected that as the loss ratio increases, or profitability
7
falls, prices would rise. The estimated sign on the coefficient might be the result of the
complex interplay between the numerator (claims costs) and the denominator (premiums)
of the loss ratio variable. As claims increase/decrease, the loss ratio rises/falls.
Inversely, as premiums increase/decrease the loss ratio would fall/rise. The negative sign
on the loss ratio variable may be capturing the premium effect dominating the claims
effect, resulting in the inverse relationship.
Equation (2) is used to test whether rate regulation is significant in explaining volatility in
the unexplained growth rate in average automobile insurance premiums. This analysis of
volatility in premium growth provides an indirect test of whether rate regulation increases
volatility in automobile insurance prices. Table 4 reports the results of this estimation.
Table 4: Coefficients and P-values for Equation (2)
Variable
Regulation
Change in average claims
Change in competition
Constant
Fixed effects
P>| t |
Coefficient
0.001
42685.9
0.805
-2.916
0.645
-18.012
0.000
15881.66
P>| t |
0.000
0.647
0.423
0.000
GLS
Coefficient
52487.7
6.525
-37.582
0.000
The index for regulation is positive and significant for both the fixed effects and GLS
models. These empirical results provide evidence that unexplained volatility in
automobile insurance premiums is higher, on average, under prior approval rate
regulation than a less regulated systems.
The lack of significance for the proxy for
opportunities for strategic behaviour (the change in the level of competition) suggests that
the scope for firm action to adapt and innovate following a change in its competitive
environment is limited.
In addition to testing whether rate regulation is important in explaining volatility in
automobile insurance premiums for all coverages, we test the effects of rate regulation for
mandatory and optional coverages and by individual line of coverage. Table 5 reports
these results.
The direction and significance of the rate regulation variable is significant and positive
for both mandatory and optional automobile coverages. This is consistent with the
practice in Canada of applying the rate regulation regime consistently to both mandatory
and optional automobile insurance purchases. In addition, the proxy for opportunities
for strategic behaviour is of the predicted sign but is not significant.
Reviewing the results by line of coverage using the fixed effects model, we find that rate
regulation is statistically significant and of the expected sign for accident benefits at the
five percent level of significance and for comprehensive coverages at the ten percent
level of significance.3
3
We also estimated these regressions using GLS but the results were similar.
8
Table 5: Rate Regulation and Volatility by Type of Coverage
Mandatory
Optional
P>| t |
0.000
0.003
Accident benefits
Third party liability
Collision
Comprehensive
0.000
0.501
0.775
0.066
Regulation
Coefficient
49132.37
43611.49
24589.99
3457.205
-322.86
-269.684
Strategic Behaviour
P>| t |
Coefficient
0.390
-26.981
0.721
-16.384
0.712
0.954
0.039
0.934
-4.115
1.624
-7.453
-0.038
The introduction of a prior approval regime in Ontario was concurrent with the
introduction of a no-fault insurance system. Under this no-fault system accident benefit
claims costs grew by 476 percent between 1989 to 2001. Such growth in claims costs
placed upward pressure on premiums and therefore the influence of rate regulation would
be expected to be important.
Rate regulation is not significant for third party liability or collision coverages. Claims
costs for third party liability and collision coverages were generally stable over the period
of the study, growing below the rate of inflation over the period.4 With less pressure on
premiums from claims costs, the effects of rate regulation would be expected to be
limited.5
An interesting result is that change in the competitive environment is a statistically
significant variable for collision coverage. One possible interpretation of this result is
that the firms in the industry have greater scope for engaging in strategic behaviour in the
repair of physical damage to vehicles as they respond to changes in the competitive
environment. As an example, firms in Ontario are able to enter into preferred collision
repair agreements to control quality and costs or have greater flexibility in responding to
consumers than is permitted in other lines of coverage.
Premium volatility in Ontario
Using the residuals obtained from equation (1) we find that the unexplained volatility in
automobile insurance prices in Ontario began increasing following the introduction of
rate regulation in Ontario in 1989. Prior to this, unexplained volatility in Ontario’s
automobile insurance premiums was less than that of Alberta and the Atlantic provinces,
which operated with file and use regulatory regimes while Ontario allowed prices to be
determined by market forces. Figure 1 shows the substantial increase in unexplained
volatility experienced in Ontario between 1989 and 2000.
4
Third party liability claims costs in Ontario experience significant volatility but on average over the
period claims cost growth was below inflation.
5
We tested whether the CPI variable was in fact a factor for these coverages and found that, while there did
appear to be some influence on the regulation variable, it did not materially alter the results.
9
Figure 1: Unexplained Volatility in Ontario
140
Ontario (historical)
120
Alberta & Atlantic Canada
Bill 59
Respond to
market
Volatilty Index
100
80
Rate regulation introduced
in Ontario
60
40
20
0
1984
1986
1988
1990
1992
1994
1996
1998
2000
The patterns depicted in Figure 1 are sharp and they provide strong evidence that rate
regulation, after accounting for changes in the environment (claims costs, accidents),
increases volatility in automobile insurance prices. It is interesting to note the change in
the slope of the volatility curve following the introduction of modifications to Ontario’s
prior approval rate regulation regime. Nevertheless, while the modified prior approval
system appears to have somewhat mitigated volatility in Ontario, it is clear from Figure 1
that it has not reduced unexplained volatility to the level of the file and use systems in the
other provinces. In addition, it is interesting to note that the slope of the curve changed
with the introduction of the ‘take-all-comers’ rule (1993) and the Bill 59 changes to the
province’s rate regulation system (1996).
While filing requirements in other provinces remain unchanged over the studied period,
rate review bodies in those provinces began asking for greater actuarial detail than they
previously required, as more detailed and sophisticated actuarial analysis became
available from solvency requirements and upgrades in insurer information technology
systems. This change in administrative practice is reflected in the increase in volatility in
those provinces following 1994. Additionally, after accounting for changes in claims
costs, accident frequency, inflation and competition, the observed trend continued into
2001 with unexplained volatility in Ontario remaining 4.93 times that of the other
provincial private insurance systems with file and use regimes.
Using our results from the analysis of rate regulation on premium volatility in the
previous section, we simulated the expected volatility for Ontario under a competitive
rating system, holding all other parameters at their historical level for the period between
1990 and 2001. Given the competitive rating system in Ontario prior to 1990, changes in
automobile expenditures were driven by changes in claims costs, accident frequency and
inflation and the province experienced very little unexplained volatility. Intuitively we
would expect to find that simulation of this system past 1989, in the absence of external
shocks to the system, would continue this trend. As expected, the simulation produced a
low and stable level of unexplained volatility after 1989.
10
Volatility and Rate Regulation: Illinois and South Carolina
The United States, with a rich heterogeneity of rate regulation across fifty states, provides
a number of natural experiments for the implications of rate regulation. We review the
experience for two illustrative United States jurisdictions, Illinois and South Carolina as
case studies with lessons for Ontario.
Table 6: Comparative Summary of Ontario, Illinois and South Carolina (2002)
(USD converted using purchasing power parity)
Population
GDP per capita
Gross Domestic Product
GDP manufacturing (%)
GDP construction( %)
GDP transportation & utilities (%)
Insured (private passenger) vehicles
Number of automobile insurers
% of vehicles in the residual market
Accident frequency per 1,000 insured vehicles
Reforms to rate/underwriting system 1989-2001
Ontario CAN$
Illinois CAN$
South Carolina CAN$
11,697,600
$37,054
$433 billion
19%
3%
5%
5,563,813
104
0.17%
43
3
12,419,293
$44,681
$555 billion
15%
5%
9%
7,190,347
386
0.03%
63
0
4,012,012
$33,279
$134 billion
21%
9%
9%
2,695,427
65 (1998)
2.7%
38
2
Differences in the automobile insurance product and other factors (legal environment etc)
mean that caution should be exercised in comparing jurisdictions. Nevertheless, the
overview of Ontario, Illinois and South Carolina (for 2000) provided in Table 6
highlights the relative similarity between Ontario and Illinois in terms of population and
the relative importance of key industrial sectors. It is interesting to note that since 1989
Ontario and South Carolina have repeatedly reformed their auto insurance rate regulation
system in response to challenges in the market, where as Illinois has been free of such
legislative activity.
Additionally, United States jurisdictions provide interesting case comparisons with
Ontario for other reasons. Specifically, in the United States there has been much
discussion and research regarding the effects of rate regulation on insurance expenditures.
Proponents of rate regulation argue that it is necessary to protect consumers from high
prices and price instability. Others argue rate regulation distorts incentives and increases
price volatility without necessarily leading to lower average prices. While it is generally
noted that claims costs are the primary drivers of changes in insurance premiums,
evidence from recent studies has supported the hypothesis that rate regulation increases
premium volatility. Anecdotal support for this idea can be found in insurance expenditure
data from Insurance Bureau of Canada and the National Association of Insurance
Commissioners (NAIC) showing (see Figure 3) price instability in the state of South
Carolina and the province of Ontario over the period of 1989 to 1999. Premium volatility
in the state of Illinois was relatively stable over the same period, with average premiums
moving with inflation.
11
Figure 3: Premium Volatility in Selected Jurisdictions
12
(2 year moving average)
M agnitude of price
movement (%)
Ontario
8
South Carolina
4
Illinois
0
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
Source: IBC and t he Nat ional Associat ion of Insurance Commissioners
Finally, the experience of South Carolina’s automobile insurance market, in terms of its
rate regulation regime and rising health care related claims costs, prior to the 1999
reform, in some respects bears resemblance to Ontario’s experience in recent years.
Therefore, South Carolina’s rate regulation reforms in 1999 provide an illustrative case
study on the consumer and market impacts for transitioning from a prior approval regime
to a competitive rating system. Illinois, a jurisdiction similar to Ontario, but which does
not regulate rates, preferred to support a competitive environment. Through a competitive
rating regime that the state has maintained for over three decades, Illinois provides an
interesting case study on the long-term implications of a competitive rating regime.
South Carolina
Similar to many other states, South Carolina introduced a prior-approval regulatory
system for auto insurance as a method of preventing insurer insolvency. More recently
the focus of rate regulation became consumer protection from rising insurance prices.
From the mid-1970s through 1998, South Carolina intensively regulated auto insurance.
Rate levels and structures were restricted, insurers’ underwriting discretion was limited,
and large cross-subsidies were channeled through its residual market (Skinner, 2003). By
1997 South Carolina suffered from a significant availability crisis as the suppression of
both voluntary and residual rates prompted insurers to exit from the state. The distortion
of economic incentives resulting from rate regulation escalated costs and prices and
caused the residual market to balloon (Shapo, 2003).
Increasing health care related claims were driving rising claims costs in the automobile
insurance system and, during the period of 1993 to 1998, claims were increasing faster
than premiums. Insurers exiting the auto insurance system outpaced entries and the
residual market held forty percent of drivers in the state. By 1996, there were only 78
companies in the auto insurance market (Grace et al, 2002).
The 1999 auto insurance reforms replaced the prior-approval system with competitive
market rating. In addition, underwriting restrictions were substantially eased as the
12
requirements for uniform classification, merit rating and rating territories were all
abolished. The residual market and its large subsidies are currently being phased out and
will be ultimately replaced by an assigned risk plan charging adequate rates (Grace et al,
2002).
Since the reforms for competitive rating were instituted, there are more insurers operating
in the state. Further, South Carolina’s ranking in terms of average premium expenditures
has improved and the residual market has decreased from 600,000 policies in 1999 to 340
policies in 2003 (Csiszar, 2003).
While most of South Carolina’s reforms became effective only in 1999, the initial impact
on the state automobile insurance market has been positive. First, the number of insurers
writing auto insurance has doubled with the implementation of the reforms. Second,
many insurers have implemented more refined risk classification and pricing structures,
as well as alternative policy options for consumers, resulting in increased choice and
greater scope for competition among insurers. Yet, while early reviews of the reform
have indicated improvement to both the affordability and availability of automobile
insurance in the state, it is still too early to draw conclusions regarding the effects of rate
regulation reforms on the volatility of premiums. Nevertheless, the South Carolina
experience provides evidence of the potentially positive effects of a competitive based
rating system.
Illinois
Since 1971, Illinois has been the only U.S. jurisdiction that has not formally regulated
automobile insurance rates. While insurance companies are required to file rate manuals,
insurance rates in the regular market are not subject to regulatory approval. Illinois
remains a distinct and interesting case because, in addition to not regulating price and
imposing few underwriting restrictions, the state does not possess the pro-market
regulation common in other jurisdictions. Illinois therefore provides an interesting case
for study because it could provide insights into whether automobile insurance rate
regulation is necessary at all.
Research on the Illinois experience has highlighted the long-run effects of a competitive
rating regime (Witt, 1977, Illinois Department of Insurance, 1979 and D’Arcy, 2002).
Since 1971, Illinois has been described as having one of the healthiest markets in the
United States with few affordability or availability issues. In general, the average
premium in Illinois has been consistently lower and experienced less volatility than the
United States average (NAII, 2003).6 Further, the state residual market has consistently
maintained a very small number of drivers, making up only one tenth of one percent of
the total market. Further, the number of insurers in Illinois is higher than the national
average providing more choice to consumers (D’Arcy, 2002).
In general, the literature has noted that due to their competitive based pricing system, the
Illinois Insurance Department has been able to focus its resources on addressing
6
According to NAIC data, Illinois was ranked 27th out of 50 states in year 2000 in terms of average
premiums
13
insurance solvency issues and monitoring market conduct. This has been described as
contributing to the benefits experienced by the Illinois auto insurance market.
Conclusion
Previous studies using U.S. data have argued that rate regulation of insurance affects both
the amplitude and length of the insurance cycle. Increased volatility in insurance
premiums is argued to be the result of delays in the rate approval process under prior
approval rate regulation. In addition, rate regulation, through higher costs of filing
requirements, also affects insurer filing and rate setting behaviour, further amplifying rate
swings caused by regulatory lag. Regulatory lags under prior approval rate regulation
could produce lower rate increases during periods of rapid cost growth and larger rate
increases or a slower rate of reduction in periods of stable or declining claims costs
(Harrington, 2002).
Our results show that the variation in insurance premiums can be largely attributable to
variation in claims costs. This is consistent with the literature based on research in the
United States supporting the hypothesis that rate regulation increases the volatility of auto
insurance premiums for consumers. As the Canadian experience with rate regulation has
been somewhat different from that of the United States, the objective of this study was to
extend this literature to the Canadian context using panel data and estimate whether rate
regulation produces increased premium volatility.
The analysis of this paper suggests that a structural shift in volatility occurred in Ontario
following the introduction of rate regulation in that province. Increased levels of rate
regulation are reliably associated with greater volatility in automobile insurance
premiums after controlling for other variables that could affect volatility. In addition, the
experiences of South Carolina and Illinois provide a natural experiment on the effects of
rate regulation, supporting the statistical evidence on the negative effects of rate
regulation and the benefits for consumers of a less regulated price system.
14
REFERENCES
Cummins, J. David and Francois Outreville. 1987. “An International Analysis of
Underwriting Cycles in Property-Liability Insurance.” Journal of Risk and Insurance 52:
16-43.
Cummins, David J., Scott E. Harrington and Robert W. Klein. 1991. “Cycles and Crises
in Property and Casualty Insurance: Causes and Implications for Public Policy,” National
Association of Insurance Commissioners, Kansas City.
Cummins, J. David, Richard Phillips and Sharon Tennyson. 2001. “Regulation, Political
Influence, and the Price of Automobile Insurance.” Journal of Insurance Regulation
20:9-50.
Csiszar, Ernst N. 2003. “The Effectiveness of State Regulation: Why Some Consumers
Cannot Obtain Insurance,” Testimony Before the Congress of the United States of
America House of Representatives Committee on Financial Services Subcommittee on
Capital Markets, Insurance and Government Sponsored Enterprises.
D’Arcy, Stephen P. 2002. “Insurance Price Deregulation: The Illinois Experience,” in
De-regulating Property-Liability Insurance: Restoring Competition and Increasing
Market Efficiency. J. David Cummins (Ed.). Washington, D.C.: AEI-Brookings Joint
Center for Regulatory Studies.
Devlin, Rose Anne. 2002. “Determinants of No-Fault Insurance Measures,” Journal of
Risk and Insurance 69:555-576.
Elay, David. 2000. “Rate and Form Regulation in the Twenty-First Century,” Journal of
Insurance Regulation 18:277-287.
Grace, Martin F., Klein, Robert W. and Phillips, Richard D. 2002. “Auto Insurance
Reform: Salvation in South Carolina,” in De-regulating Property-Liability Insurance:
Restoring Competition and Increasing Market Efficiency. J. David Cummins (ed.).
Washington, D.C.: AEI-Brookings Joint Center for Regulatory Studies.
Harrington, Scott E. 1984. “A Note on the Impact of Auto Insurance Rate Regulation.”
Review of Economics and Statistics 69: 166-170.
Harrington, Scott E. and Greg Niehaus. 2000. “Volatility and Underwriting Cycles.” in
Handbook of Insurance, Georges Dionne, ed., Huebner International Series on Risk,
Insurance and Economic Security.
Harrington, Scott E. 2001. “Effects of Prior Approval Regulation of Automobile
Insurance” in De-regulating Property-Liability Insurance: Restoring Competition and
Increasing Market Efficiency, J. David Cummins (ed.), Washington, D.C.: AEIBrookings Joint Center for Regulatory Studies.
Illinois Department of Insurance. 1977. “Illinois Automobile Rate Study,” Springfield.
15
Kovacs, Paul and Darrell Leadbetter. 2004. “The Economic Burden of Motor Vehicle
Collisions in British Columbia: The Influence of Social Pricing in Insurance,” Insurance
and Risk Management Journal (forthcoming).
Lamm-Tennant, Joan and Mary A. Weiss. 1997. “International Insurance Cycles:
Rational Expectations/International Intervention,” Journal of Risk and Insurance 64: 415439.
Leng, Chao-chun, Michael R. Powers and Emilio C. Venezian. 2002. “Did Regulation
Change Competitiveness in Property-Liability Insurance? Evidence from Underwriting
and Investment Income,” Journal of Insurance Regulation 21: 57-78.
Miller, Harry and Robert Witt. “Price Competition, Regulation and Systemic
Underwriting Risk in Automobile Insurance Markets,” CPCU Journal 34:174-89.
NAII. 2003. “Industry Issues: Regulatory Modernization,” National Association of
Independent Insurers http://www.naii.org/Sitehome.nsf/RegulatoryModernizationBackgrounder.
Outreville, Francois J. 1990. “Underwriting Cycles and Rate Regulation in Automobile
Insurance Markets,” Journal of Insurance Regulation 8:274-286.
Shapo, Nathaniel S., 2003. “Shopping for a Solution: Effective Consumer Protection
Through Competitive Regulation of Insurance Rates,” Texas Public Policy Foundation,
Austin, Texas.
Skinner, Brett. 2003. “The Political Context for Changes in State Automobile Insurance
Regulation: A Policy Brief,” Insurance Bureau of Canada working paper.
Tennyson, Sharon. 1991. “The Effect of Rate Regulation on Underwriting Cycles.”
CPCU Journal 44(1):33-45.
Tennyson, Sharon. 1997. “The Impact of Rate Regulation on State Automobile
Insurance Markets.” Journal of Insurance Regulation 15:502-523.
Toyoda, T. (1974), “Use of the Chow Test Under Heteroskedasticity,” Econometrica, 42,
3 (May), pp.601-608
Witt, Robert C. 1977. “The Automobile Insurance Rate Regulatory System in Illinois: A
Comparative Study,” Illinois Insurance Laws Study Commission.
16